Original Paper. Time Since Last Use of Oral Contraceptives and Risk of Invasive Cervical Cancer

Transcription

1 Pergamon European Journal of Cancer, Vol. 34, No. 6, pp. 884±888, 1998 # 1998 Elsevier Science Ltd. All rights reserved Printed in Great Britain /98 $ PII: S (97) Original Paper Time Since Last Use of Oral Contraceptives and Risk of Invasive Cervical Cancer F. Parazzini, 1,2 L. Chatenoud, 1 C. La Vecchia, 1,3 F. ChiaVarino, 1 E. Ricci 1 and E. Negri 1 1 Istituto di Ricerche Farmacologiche `Mario Negri', via Eritrea 62, Milan; 2 Prima Clinica Ostetrico Ginecologica, Divisione di Oncologia Ginecologica, Istituto Nazionale Tumori, UniversitaÁ di Milano, via Commenda 12, Milan; and 3 Istituto di Statistica Medica e Biometria, UniversitaÁ di Milano, via Venezian 1, Milan, Italy The time±risk relationship for the association between cervical cancer and oral contraceptives (OC) was examined using data on 592 cases of invasive cervical cancer aged 60 years or less and 616 controls with acute, non-gynaecological, non-hormone-related, non-neoplastic diseases. A total of 125 cases and 114 controls reported ever using OC and the multivariate odds ratio (OR) for ever versus never users was 1.21 (95% con dence interval (CI) 0.82±1.74). The risk of invasive cervical cancer was above unity in current users (OR 1.23) and in women who had stopped OC use less than 10 years before diagnosis, but not in those who had stopped their OC use 10 years before (OR 0.85). Similarly, the OR was less for women who had started OC use 15 years or more previously than for more recent users. These data suggest that OCs may have a late stage (promoter) evect on cervical carcinogenesis and thus have public health implications, since the incidence of invasive cervical cancers is low at young ages, when OC use is more common and increases during middle age. The absence of a persisting risk is therefore of interest both for assessing individual risk and for its public health implications. # 1998 Elsevier Science Ltd. All rights reserved. Key words: cervical cancer, oral contraceptives, risk factors Eur J Cancer, Vol. 34, No. 6, pp. 884±888, 1998 INTRODUCTION Several epidemiological studies have suggested that parous women and oral contraceptive (OC) users are at increased risk of invasive cervical cancer [1±4]. In biological terms, it has been postulated that oestrogen±progestin stimulation can favour, or promote, cervical carcinogenesis, possibly through glucocorticoid-dependent oncogenic transformation by selected papilloma virus types [5]. This suggests that hormonal stimulation may act on one of the latter stages of the process of cervical carcinogenesis. A detailed analysis of the risk relationship over time of any association between cervical cancer and OCs may, therefore, help elucidate the role of hormonal correlates of cervical carcinogenesis, as well as de ne the bene t/risk pro le of OCs. For example, an update of the Oxford Family Planning Correspondence to F. Parazzini. Received 25 Jul. 1997; revised 5 Nov. 1997; accepted 3 Dec Association cohort study indicated that the elevated risk of cervical cancer associated with OC use was essentially con- ned to current or recent long-term users [6]. To provide further information on these issues, we considered data collected in a large case±control study of cervical neoplasia conducted in Italy. MATERIALS AND METHODS This was a hospital-based case±control investigation conducted in the greater Milan area, northern Italy between 1981 and Its general design has been described in previous papers including a subset of subjects considered in this report [4, 7]. Cases were 592 histologically con rmed invasive cervical cancers below the age of 61 years (range 17±60 years, median 48 years) admitted to a network of Obstetrics and Gynaecology Departments in the great Milan area. The comparison group consisted of 616 women aged 60 years or less (range 16±60 years, median 48 years), 884

2 Oral Contraceptives and Cervical Cancer 885 Table 1. Distribution of 592 invasive cervical cancer cases and 616 controls according to age and selected factors, Milan, Italy 1981±1993 Number of cases (%) Number of controls (%) Multivariate odds ratio (95% CI)* Age (years) < (25.5) 170 (27.6) ± 40± (30.7) 161 (26.1) ± 50± (43.7) 285 (46.2) ± Education (years) < (63.0) 310 (50.3) 1y 7± (21.5) 185 (30.0) 0.71 (0.53±0.98) (15.5) 121 (19.6) 0.78 (0.54±1.16) Smoking habits Never smokers 380 (64.2) 398 (64.6) 1y Ever smokers 212 (35.8) 218 (35.4) 1.06 (0.80±1.42) Parity 0 54 (9.1) 130 (21.1) 1y 1±2 310 (52.4) 346 (56.2) 3.11 (2.08±4.69) (38.5) 140 (22.7) 5.42 (3.48±8.56) Lifelong number of sexual partners 0 4 (0.7) 35 (5.7) 0.22 (0.07±0.67) (69.8) 457 (74.2) 1y 2 81 (13.7) 62 (10.1) 1.80 (1.20±2.70) 3 86 (14.5) 49 (7.9) 3.85 (1.91±4.32) Unreported 8 (1.4) 13 (2.1) ± Lifelong number of normal Pap smears (57.1) 161 (26.1) 1y 1 90 (15.2) 89 (14.4) 0.39 (0.27±0.57) 2 91 (15.4) 63 (10.2) 0.26 (0.16±0.41) 3 70 (11.8) 296 (48.1) 0.15 (0.11±0.21) Unreported 3 (0.5) 7 (1.1) Abnormal Pap smearsz Never 240 (94.5) 441 (96.9) 1y Ever 14 (5.5) 14 (3.1) 1.49 (0.61±3.20) CI, con dence interval. *Multivariate estimates including terms for age, education, calendar year at interview, smoking habits, parity, lifelong number of sexual partners and of Pap smears, and oral contraceptive use. yreference categories. zone year before interview or more. Table 2. Distribution of 592 invasive cervical cancer cases and 616 controls according to selected measures of oral contraceptive (OC) use, corresponding odds ratios (OR) and 95% con dence intervals (CI), Milan, Italy 1981±1993 OR (95% CI)* Number of cases (%) Number of controls (%) Age-adjusted Multivariate OC use Ever use 125 (21.1) 114 (18.5) 1.21 (0.89±1.64) 1.21 (0.82±1.74) Duration of OC use (months) (10.5) 56 (9.1) 1.23 (0.82±1.83) 1.17 (0.75±1.82) > (10.6) 56 (9.1) 1.32 (0.81±2.15) 1.31 (0.72±2.39) Unknown ± 2 (0.4) ± ± w 2 1 trend 1.33 (P = 0.25) 3.05 (P = 0.08) Time since last use (years) Current 19 (3.2) 23 (3.7) 1.01 (0.53±1.94) 1.33 (0.65±2.70) 1± < 3 23 (3.9) 17 (2.8) 1.61 (0.82±3.13) 1.56 (0.74±3.29) 3± < 6 22 (3.7) 16 (2.6) 1.60 (0.81±3.17) 1.47 (0.69±3.14) 6± < (4.4) 17 (2.8) 1.63 (0.86±3.09) 1.66 (0.83±3.34) (5.9) 41 (6.7) 0.90 (0.56±6.45) 0.85 (0.51±1.43) Time since rst use (years) 5 26 (4.4) 25 (4.1) 1.33 (0.72±2.45) 1.49 (0.76±2.94) 6± < (6.1) 22 (3.6) 1.89 (1.06±3.35) 1.77 (0.93±3.37) 11± < (5.1) 32 (5.2) 1.02 (0.60±1.74) 1.07 (0.60±1.90) (5.6) 35 (5.7) 0.98 (0.60±1.61) 0.96 (0.55±1.65) Age at rst OC use < (6.4) 40 (6.5) 1.12 (0.68±1.86) 1.12 (0.63±1.97) 25±29 35 (5.9) 35 (5.7) 1.10 (0.66±1.83) 1.10 (0.63±1.92) (8.8) 39 (6.3) 1.35 (0.87±2.10) 1.36 (0.83±2.23) *Multivariate estimates including terms for age, education, calendar year at interview, parity, number of sexual partners, smoking habits and number of Pap smears. yreference category.

3 886 F. Parazzini et al. with acute, non-gynaecological, non-hormone-related, nonneoplastic diseases and who had not undergone total hysterectomy, admitted to the same network of hospitals where the cases had been identi ed, and from comparable catchment areas. Of these, 26% were admitted for traumatic conditions (mostly fractures and sprains), 35% had nontraumatic orthopaedic disorders (mostly lower back pain and disc disorders), 14% surgical conditions (mostly abdominal, such as acute appendicitis or strangulated hernia) and 24% had other illnesses, such as ear, nose and throat or dental disorders. Less than 2% of eligible women (cases and controls) refused to be interviewed. A structured questionnaire was used to collect information on personal characteristics and habits, education and other socio-economic factors, general lifestyle habits, such as smoking, alcohol and covee consumption, frequency of consumption of a few indicator foods, selected indicators of sexual habits, menstrual and reproductive history, related medical history, and history of lifelong use of OCs and female hormone preparations for any other indication, including time and duration of each episode of use and the brand name, whenever available. In addition, information on the lifetime number of Pap smears was collected. Odds ratios (OR) of cervical cancer, and the corresponding 95% con dence intervals (CI) for various measures of OC use were derived using unconditional multiple logistic regression, tted by the method of maximum likelihood [8], including terms for age in quinquennia, education ( < 7, 7±11, 12 years), calendar year and interview, smoking habits (ever, never), parity (0, 1±2, 3), number of partners (0, 1, 2) and of Pap smears (0, 1, 2, 3). RESULTS Table 1 shows the distribution of cases and controls according to age and selected covariates. Cases were less educated, more frequently multiparous, reported a lifelong larger number of sexual partners and less frequently underwent Pap smears. In Table 2, diverent measures of OC use are considered. A total of 125 cases and 114 controls reported ever using OCs. The corresponding OR was 1.21 (95% CI 0.82±1.74, multivariate). The ORs were 1.17 in users for 24 months or less and 1.31 in users for > 24 months (multivariate; w 2 1 trend 3.05, P = 0.08). The risk of invasive cervical cancer was above unity for current users or in women who had stopped use less than 10 years before they were interviewed, but not in those who stopped their OC use 10 years before, whose OR was 0.85 (95% CI 0.51±1.43, multivariate). Likewise, in comparison with never users, the OR was 0.96 (95% CI 0.55±1.65, multivariate) for women who had started OC use 15 or more years previously. There was no clear relationship between age at rst OC use and risk of cervical cancer. Table 3 shows the analyses of the relationship between the time since last OC use and the risk of invasive cervical cancer in strata of selected covariates. The ORs were systematically higher in current OC users or ex-users from less than 10 years before, than in those who had stopped OC use for 10 years or more. The association with recent use was apparently stronger in less educated, multiparous women, who had never undergone a Pap smear (i.e. at baseline high risk). For example, the OR of cervical cancer in women reporting the time since last OC use as < 10 years was 2.22 in women Table 3. Multivariate odds ratios (ORs)* of invasive epithelial cervical cancer [and corresponding con dence intervals (CI)] according to time since last oral contraceptive (OC) use in strata of selected covariates, Milan, Italy 1981±1993 Time since last OC use Never OC use Current < 10 years 10 years Total 1y 1.23 [0.60±2.44]z 1.57 [0.99±2.49] 0.85 [0.51±1.43] (19:23) (71:50) (35:41) Age (years) < 45 1y 1.26 [0.59±2.71] 1.42 [0.83±2.41] 0.67 [0.29±1.55] (17:21) (60:42) (14:18) 45±60 1y 0.64 [0.08±5.07] 1.53 [0.55±4.23] 0.99 [0.51±1.93] (2:2) (11:8) (21:23) Number of sexual partners 0±1 1y 1.38 [0.49±3.87] 1.41 [0.80±2.50] 1.10 [0.59±2.05] (8:11) (36:33) (26:26) 2 1y 0.94 [0.35±2.47] 1.74 [0.80±3.79] 0.51 [0.19±1.60] (11:12) (35:17) (9:15) Education (years) < 7 1y 0.70 [0.16±2.99] 2.22 [1.10±4.50] 1.13 [0.54±2.35] (4:6) (39:17) (21:18) 7 1y 1.24 [0.55±2.80] 1.12 [0.60±2.09] 0.62 [0.27±1.37] (15:17) (32:33) (14:23) Parity 0±1 1y 1.25 [0.44±2.51] 1.24 [0.61±2.54] 1.10 [0.44±3.07] (13:15) (20:27) (9:12) > 2 1y 0.95 [0.30±3.07] 1.97 [1.07±3.62] 0.89 [0.48±1.66] (6:8) (51:23) (26:29) Lifetime number of Pap smears 0±1 1y 0.65 [0.22±1.86] 2.30 [1.06±4.97] 1.32 [0.56±3.07] (8:10) (45:10) (24:8) 2 1y 1.28 [0.50±3.29] 1.10 [0.60±2.01] 0.59 [0.28±1.24] (10:12) (26:40) (11:33) *Multivariate estimates including terms for age, education, calendar year at interview, parity, number of partners, smoking habits and number of Pap smears. yreference category. z95% con dence interval.

4 Oral Contraceptives and Cervical Cancer 887 Table 4. Main results from selected studies on recency and duration of oral contraceptive use and risk of invasive cervical cancer Author, year [ref.] Type of study Number of cases Years of use Relative risk* Years since last use Relative risk* Peters and associates, 1986 [11] Brinton and associates, 1986 [12] Ebeling and associates, 1987 [13] Irwin and associates, 1988 [14] Beral and associates, 1988 [15] Cuzick and associates, 1989 [16] Brinton and associates, 1990 [17] Bosch and associates, 1992 [18] Case±control 200 < ± ± 2± Case±control 479 < ±9 2.0 > ± ± Case±control 129 1±3 1.1 Previous use 1.2 4±6 1.2 Present use Case-control 149 < < ± ±4 1.0 Cohort 65 < y 5± Case±control 135 < (including 55 CIN III, 1± micro-invasive, > 3± invasive cancers) > Case±control 667 < ±9 1.3 > Case±control 436 1±9z ±9x WHO, 1993 [19] Case±control < ±1 1.2 > 1±3 1.6 > 1±2 1.1 > 3±8 1.2 > 2±3 1.2 > 8± > 3±5 1.4 > > 5±8 1.5 > Eluf-Neto and associates, 1994 [20] Zonderwan and associates, 1996 [6] (Oxford FPA) Case±control 199 1± Cohort 33 1± > 2±6 2.8 > 2±6 2.6 > > *Reference category: never users. yrate ratio. zhuman papillomavirus negative cases. xhuman papillomavirus positive cases. reporting < 7 years of education and 2.30 in those reporting 0±1 lifetime Pap smears. DISCUSSION In this study, the risk of invasive epithelial cervical cancer tended to be elevated in ever-users of OC and to increase with duration of use, but the trend in risk was not signi cant. The risk seems to be similar to that among current users, up to 10 years since last use, before decreasing for women who had last used OCs more than 10 years ago. These ndings agree with the suggestion [6] that the excess risk of cervical cancer is largely restricted to current or recent long-term OC users and tends to decline with time since last use. Consequently, the present data suggest that the evect of oestroprogestins would be on one of the late stages of the process of cervical carcinogenesis (i.e. promotion). Potential limitations of this study should be considered. OC users may tend to be more frequently screened by Pap smear [9]. This potential selection mechanism should lead to an underestimate of the role of OCs, particularly in the short term and in the present analysis the evect of OCs on cervical cancer risk was larger in women who did not report any Pap smear during their life. With regard to other covariates (chie y education and sexual habits), their inclusion in the multivariate analysis did not drastically change any of the ORs. With reference to information bias, the similar clinical setting should favour a comparable recalling of drug use (including OCs) among cases and controls [10]. The questions regarding OC use were derived from a questionnaire including more than 100 questions, analysing general characteristics, reproductive and medical history and general lifestyle habits. Further, the analysis of the relationship between OC use and cervical cancer risk was not the main objective of the study and the interviewers, as well as the women interviewed, were not aware of the potential association between OC use and the risk of invasive cervical cancer. This association, moreover, had not gained widespread attention in the lay press in Italy. Selection bias should also not be a major problem in the study, since cases and controls were identi ed in institutions covering broadly comparable catchment areas and participation was almost complete. It is unlikely that any possible source of bias had speci cally in uenced the time±risk relationship of any OC±cervical cancer risk.

5 888 F. Parazzini et al. The results of the study are in general agreement with a recent update of the Oxford Family Planning Association contraceptive study [6], in that ever users of OCs had an elevated OR of cervical invasive neoplasm. In that study, the separate evects of time since last use and duration of use were examined and the risk was largely con ned to current or recent ( < 2 years), long-term users of OCs. The potential confounding by number of sexual partners and of cervical smears could not be taken into account, but it was thought that this was unlikely to have introduced major bias. Similar ndings also emerged in other studies. Table 4 shows the results of selected published papers on the relationship between recency and duration of OC use and risk of invasive cervical cancer. In biological terms, the evect of OC use in cervical carcinogenesis may be attributed to the progestin component of the pill. HPV-16, the most important cause of cervical cancer, contains a progesterone/glucocorticoid response element upstream to the common E6-E7 promoter [21] and progesterone enhances the transferring evect of HPV-16 DNA [22]. In fact, papillomavirus lesions are exacerbated during pregnancy [23], and cervical cancer risk is enhanced by OCs and pregnancy, when progestogen levels are relatively high. Otherwise, oestrogen (prolonged by local 16a-hydroxylation) increases human papillomavirus expression via upregulation of the progesterone receptor [24]. This line of reasoning indicates that, in terms of clinical and public health implications, OC use should be critically reconsidered for women with a diagnosis of cervical intraepithelial neoplasia. However, little association between OC use and cervical cancer emerged in screened women, suggesting that, from a public health perspective, any potential impact of OC on cervical cancer risk use would be largely controlled by Pap screening. The major public health implications of this study, however, are the reassuring evidence on the absence of persisting cervical cancer risk in the medium±long term after stopping OC use. This is of particular interest, since the incidence of invasive cervical cancer is low at a younger age (i.e. below age 35 years), when OC use is most common and increases throughout middle age. Speci c attention to screening for cervical neoplasia, in contrast, would be recommended for OC users above the age of 40 or 45 years. 1. Brinton LA. Epidemiology of cervical cancerðoverview. In MunÄoz N, Bosch FX, Shah KV, Meheus. A, eds. The Epidemiology of Cervical Cancer and Human Papillomavirus. Lyon, International Agency for Research on Cancer IARC, 1992, 3± Brinton LA, Hamman RF, Huggins GR, et al. Sexual and reproductive risk factors for invasive squamous cell cervical cancer. J Natl Cancer Inst 1987, 79, 23± Hidesheim A, Reeves WC, Brinton LA, et al. Association of oral contraceptive use and human papillomaviruses in invasive cervical cancer. Int J Cancer 1990, 45, 860± Parazzini F, La Vecchia C, Negri E, Cecchetti G, Fedele L. Reproductive factors and the risk of invasive and intra-epithelial cervical neoplasia. Br J Cancer 1989, 59, 805± Pater MM, Hughes GA, Hyslop DE, Nakshatri H, Pater A. Glucocorticoid-dependent oncogenic transformation by type 16 but not type 11 human papilloma virus DNA. Nature 1988, 335, 832± Zondervan KT, Carpenter LM, Painter R, Vessey MP. Oral contraceptives and cervical cancerðfurther ndings from the Oxford Family Planning Association contraceptive study. Br J Cancer 1996, 73, 1291± Parazzini F, La Vecchia C, Negri E, Maggi R. Oral contraceptive use and invasive cervical cancer. Int J Epidemiol 1990, 19, 259± Baker NJ, Nelder JA. The GLIM System, Release 3. Oxford, Numerical Algorithms Group, Parazzini F, Negri E, Ricci E, Franceschi S, La Vecchia C. Correlates of oral contraceptive use in Italian women, 1991±93. Contraception 1996, 54, 101± Paganini-Hill A, Ross RK. Reliability of recall of drug usage and other health-related information. Am J Epidemiol 1982, 116, 114± Peters RK, Thomas D, Hagan DG, Mack TM, Henderson BE. Risk factors for invasive cervical cancer among latinas and nonlatinas in Los Angeles county. J Natl Cancer Inst 1986, 77, 1063± Brinton LA, Huggins GR, Lehman HF, et al. Long-term use of oral contraceptives and risk of invasive cervical cancer. Int J Cancer 1986, 38, 339± Ebeling K, Nischan P, Schindler Ch. Use of oral contraceptives and risk of invasive cervical cancer in previously screened women. Int J Cancer 1987, 39, 427± Irwin KL, Rosero-Bixby L, Oberle MW, et al. Oral contraceptives and cervical cancer risk in Costa Rica. J Am Med Assoc 1988, 259, 59± Beral V, Hannaford P, Kay C. Oral contraceptive use and malignancies of the genital tract. Lancet 1988, 127, Cuzick J, De Stavola B, McCance D, et al. A case±control study of cervix cancer in Singapore. Br J Cancer 1989, 60, 238± Brinton LA, Reeves WC, Brenes MM, et al. Oral contraceptive use and risk of invasive cervical cancer. Int J Epidemiol 1990, 19, 4± Bosch FX, Munoz N, De Sanjose S, et al. Risk factors for cervical cancer in Colombia and Spain. Int J Cancer 1992, 52, 750± WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Invasive squamous-cell cervical carcinoma and combined oral contraceptives: results from a multinational study. Int J Cancer 1993, 55, 228± Eluf-Neto J, Booth M, Munoz N, Bosch FX, Meijer CJLM, Walboomers JMM. Human papillomavirus and invasive cervical cancer in Brazil. Br T Cancer 1994, 69, 114± Chan WK, Klock G, Bernard HV. Progesterone and glucocorticoid response elements occur in the long control regions of several human papillomaviruses involved in anogenital neoplasia. J Virol 1989, 63, 3261± Pater A, Bayatiour M, Pater MM. Oncogenic transformation by human papillomavirus type 16 deoxyribonucleic acid in the presence of progesterone or progestins from oral contraceptives. Am J Obstet Gynec 1990, 162, 1099± Schneider A, Hotz M, Gissmann I. Increased prevalence of human papillomaviruses in the lower genital tract of pregnant women. Int J Cancer 1987, 40, 198± Monsonego J, Magdelenat H, Catalan F, Coscas Y, Zerat L, Sastre X. Estrogen and progesterone receptors in cervical human papillomavirus related lesions. Int J Cancer 1991, 48, 533±539. AcknowledgementsÐThis work was conducted within the framework of the CNR (Italian National Research Council) Applied Projects `Clinical Applications of Oncological Research' (Contract no PF39) and `Risk Factors for Disease' (Contract no PF41), and with the contribution of the Italian Association for Cancer Research and the Italian League against Tumours. The authors with to thank Ms Ivana Garimoldi for editorial assistance.